Batch text generation

Batch predictions are a way to efficiently send multiple multimodal prompts that are not latency sensitive. Unlike online prediction, where you are limited to one input prompt at a time, you can send a large number of multimodal prompts in a single batch request. Then, your responses asynchronously populate in your BigQuery storage output location.

Batch requests for Gemini models are discounted 50% from standard requests. To learn more, see the Pricing page.

Multimodal models that support batch predictions

The following multimodal models support batch predictions.

  • gemini-1.5-flash-001
  • gemini-1.5-pro-001
  • gemini-1.0-pro-002
  • gemini-1.0-pro-001

Prepare your inputs

Batch requests for multimodal models accept BigQuery storage sources and Cloud Storage sources.

BigQuery storage input

  • The content in the request column must be valid JSON. This JSON data represents your input for the model.
  • The content in the JSON instructions must match the structure of a GenerateContentRequest.
  • Your input table can have columns other than request. They are ignored for content generation but included in the output table. The system reserves two column names for output: response and status. These are used to provide information about the outcome of the batch prediction job.
  • Batch prediction doesn't support the fileData field for Gemini.
Example input (JSON) 
        
{
  "contents": [
    {
      "role": "user",
      "parts": {
        "text": "Give me a recipe for banana bread."
      }
    }
  ],
  "system_instruction": {
    "parts": [
      {
        "text": "You are a chef."
      }
    ]
  }
}

Cloud Storage input

  • File format: JSON Lines (JSONL)
  • Located in us-central1
  • Appropriate read permissions for the service account
  • fileData limitations for certain Gemini models.
    Example input (JSONL) 
    
{"request":{"contents": [{"role": "user", "parts": [{"text": "What is the relation between the following video and image samples?"}, {"file_data": {"file_uri": "gs://cloud-samples-data/generative-ai/video/animals.mp4", "mime_type": "video/mp4"}}, {"file_data": {"file_uri": "gs://cloud-samples-data/generative-ai/image/cricket.jpeg", "mime_type": "image/jpeg"}}]}]}}
{"request":{"contents": [{"role": "user", "parts": [{"text": "Describe what is happening in this video."}, {"file_data": {"file_uri": "gs://cloud-samples-data/generative-ai/video/another_video.mov", "mime_type": "video/mov"}}]}]}}

Request a batch response

Depending on the number of input items that you submitted, a batch generation task can take some time to complete.

REST

To test a multimodal prompt by using the Vertex AI API, send a POST request to the publisher model endpoint.

Before using any of the request data, make the following replacements:

  • PROJECT_ID: The name of your Google Cloud project.
  • BP_JOB_NAME: A name you choose for your job.
  • INPUT_URI: The input source URI. This is a BigQuery table URI in the form bq://PROJECT_ID.DATASET.TABLE. Or your Cloud Storage bucket URI.
  • INPUT_SOURCE: The input source type. Options are bigquerySource and gcsSource.
  • INSTANCES_FORMAT: Input instances format - can be `jsonl` or `bigquery`.
  • OUTPUT_URI: The URI of the output or target output table, in the form bq://PROJECT_ID.DATASET.TABLE. If the table doesn't already exist, then it is created for you.

HTTP method and URL: 

POST https://us-central1-aiplatform.googleapis.com/v1/projects/PROJECT_ID/locations/us-central1/batchPredictionJobs

Request JSON body: 

{
    "displayName": "BP_JOB_NAME",
    "model": "publishers/google/models/gemini-1.0-pro-002",
    "inputConfig": {
      "instancesFormat":"INSTANCES_FORMAT",
      "inputSource":{ INPUT_SOURCE
        "inputUri" : "INPUT_URI"
      }
    },
    "outputConfig": {
      "predictionsFormat":"bigquery",
      "bigqueryDestination":{
        "outputUri": "OUTPUT_URI"
        }
    }
}

To send your request, choose one of these options:

curl

Save the request body in a file named request.json, and execute the following command: 

curl -X POST \
     -H "Authorization: Bearer $(gcloud auth print-access-token)" \
     -H "Content-Type: application/json; charset=utf-8" \
     -d @request.json \
     "https://us-central1-aiplatform.googleapis.com/v1/projects/PROJECT_ID/locations/us-central1/batchPredictionJobs"

PowerShell

Save the request body in a file named request.json, and execute the following command: 

$cred = gcloud auth print-access-token
$headers = @{ "Authorization" = "Bearer $cred" }

Invoke-WebRequest `
    -Method POST `
    -Headers $headers `
    -ContentType: "application/json; charset=utf-8" `
    -InFile request.json `
    -Uri "https://us-central1-aiplatform.googleapis.com/v1/projects/PROJECT_ID/locations/us-central1/batchPredictionJobs" | Select-Object -Expand Content

You should receive a JSON response similar to the following:

{
  "name": "projects/{PROJECT_ID}/locations/us-central1/batchPredictionJobs/{BATCH_JOB_ID}",
  "displayName": "My first batch prediction",
  "model": "projects/{PROJECT_ID}/locations/us-central1/models/gemini-1.0-pro-002",
  "inputConfig": {
    "instancesFormat": "bigquery",
    "bigquerySource": {
      "inputUri": "bq://{PROJECT_ID}.mydataset.batch_predictions_input"
    }
  },
  "modelParameters": {},
  "outputConfig": {
    "predictionsFormat": "bigquery",
    "bigqueryDestination": {
      "outputUri": "bq://{PROJECT_ID}.mydataset.batch_predictions_output"
    }
  },
  "state": "JOB_STATE_PENDING",
  "createTime": "2023-07-12T20:46:52.148717Z",
  "updateTime": "2023-07-12T20:46:52.148717Z",
  "modelVersionId": "1"
}

The response includes a unique identifier for the batch job. You can poll for the status of the batch job using the BATCH_JOB_ID until the job state is JOB_STATE_SUCCEEDED. For example:

curl \
  -X GET \
  -H "Authorization: Bearer $(gcloud auth print-access-token)" \
  -H "Content-Type: application/json" \
https://us-central1-aiplatform.googleapis.com/v1/projects/PROJECT_ID/locations/us-central1/batchPredictionJobs/BATCH_JOB_ID

Retrieve batch output

When a batch prediction task completes, the output is stored in the BigQuery table that you specified in your request.

BigQuery output example

request response status
'{"content":[{...}]}' 
{
  "candidates": [
    {
      "content": {
        "role": "model",
        "parts": [
          {
            "text": "In a medium bowl, whisk together the flour, baking soda, baking powder."
          }
        ]
      },
      "finishReason": "STOP",
      "safetyRatings": [
        {
          "category": "HARM_CATEGORY_SEXUALLY_EXPLICIT",
          "probability": "NEGLIGIBLE",
          "probabilityScore": 0.14057204,
          "severity": "HARM_SEVERITY_NEGLIGIBLE",
          "severityScore": 0.14270912
        }
      ]
    }
  ],
  "usageMetadata": {
    "promptTokenCount": 8,
    "candidatesTokenCount": 396,
    "totalTokenCount": 404
  }
}

Cloud Storage output example

PROJECT_ID=[PROJECT ID]
REGION="us-central1"
MODEL_URI="publishers/google/models/gemini-1.0-pro-001@default"
INPUT_URI="[GCS INPUT URI]"
OUTPUT_URI="[OUTPUT URI]"

# Setting variables based on parameters
ENDPOINT="${REGION}-autopush-aiplatform.sandbox.googleapis.com"
API_VERSION=v1
ENV=autopush
BP_JOB_NAME="BP_testing_`date +%Y%m%d_%H%M%S`"

curl \
  -X POST \
  -H "Authorization: Bearer $(gcloud auth print-access-token)" \
  -H "Content-Type: application/json" \
  https://${ENDPOINT}/${API_VERSION}/projects/${PROJECT_ID}/locations/${REGION}/batchPredictionJobs \
-d '{
    "name": "'${BP_JOB_NAME}'",
    "displayName": "'${BP_JOB_NAME}'",
    "model": "'${MODEL_URI}'",
    "inputConfig": {
      "instancesFormat":"jsonl",
      "gcsSource":{
        "uris" : "'${INPUT_URI}'"
      }
    },
    "outputConfig": {
      "predictionsFormat":"jsonl",
      "gcsDestination":{
        "outputUriPrefix": "'${OUTPUT_URI}'"
      }
    },
    "labels": {"stage": "'${ENV}'"},
}'

What's next